Conserved emergent traits enable biobank-scale prediction of community function

A central challenge in microbial ecology and biotechnology is predicting how species identity translates into community-level function. Using soymilk fermentation as a model, we measured three industrially relevant functions - acidification, texture, sensory grade - across 307 synthetic communities combining 33 phylogenetically diverse lactic acid bacteria strains. The functional effect of adding a strain to a community scaled linearly with the function of the receiving community. This global epistasis-like pattern summarizes each strains contribution across communities with two emergent parameters - intercept and slope - which hold predictive power across untested combinations. We show that these parameters are phylogenetically conserved, allowing their imputation from 16S rRNA identity alone. This scales up the predictive capabilities of our method to communities composed of entirely unassayed strains. By redefining community-level function in terms of conserved species traits, our results pave the way for biobank-scale consortium engineering and genomic dissection of complex community phenotypes.